The basic principles of algorithm based on the LOG operator’s local dynamic threshold and the maximum between clusters variance’s global thresholds are respectively analyzed. The specific algorithm of binarization is detailedly discussed in digital recognition of mechanical workpiece. For their advantages of dynamic threshold and global threshold and in view of the characteristics of workpiece character, a new method is proposed and combines the local dynamic threshold grounded on the LOG operator with the global threshold of maximum between clusters variance. To begin with, the confusing and resemblant figures characters can be easily distinguished by using local dynamic threshold algorithm of LOG operator to enhance the details of characters. And then the binarization of image background is executed by use of the global threshold algorithm of maximum between class clusters. The integrated algorithm has good performance in many aspects such as computational complexity, implementation time, binarization effects and the scope of application. It has been applied to the actual system and achieves good results.

The velocity measurement accuracy 1% from 4.06 ~475.3 mm/s and analysis of accuracy for Laser Doppler Velocimers (LDVs) based on self-mixing effect are reported. We specifically address the study of broadening of Doppler frequency due to laser spectrum, velocity distribution, speckle pattern and electronics. It can be concluded that the velocity measurement error is mainly caused by speckle patterns especially with small velocity. Extension of sampling time and selecting proper signal processing circuits is a good solution for improving velocity measurement accuracy.

To get the 6D parameters of an object, two cameras are at least involved in the system to shoot at different positions so that the stereopair with a certain parallax can be obtained, and then the space parameters of the object will be worked out through space intersection. By changing the classical photogrammetric collinearity equations, parameters to be solved are taken as equation parameters to obtain extended collinearity equations. When the rigid object have three or more reference marks of known coordinates, the 6-dimensional space parameters of target can be got by use of single frame images. Because of fewer cameras to be needed, the technique can be applied to the occasions that too many cameras are not allowed to be installed , such as flight testing, aircraft components docking.

Considering the problem that the image of surface mounting component captured from automated optical inspection system exists parallel motion and circumrotatory error, a method for capturing accurate image based on subpixel location is proposed. Combined with Canny edge detection, based on the subpixel location of reference position, the accurate image of surface mounting component is captured. It overcomes the position error and manufacturing error of other circuit boards and improves the accuracy of detection for automated optical inspection system. The experiment proves that the image of component can be captured with high accuracy and robustness, which can reach the requirement of measurement precision.

Traditional particle filter tracking algorithm is difficult to get accurate target location in the complicated circumstance. In order to solve these problems, a novel particle filter tracking algorithm based on online feature selection is proposed. Firstly, the algorithm converted the input image into sixteen different color feature spaces, and fisher discriminating rule was adopted to select the top-ranked feature space which could discriminate the target region and the neighbor background region best. Then, particle filter algorithm based on statistical histogram was applied to track object in the top-ranked feature space. Experimental results show that this algorithm is robust and the target is tracked accurately under the conditions of illumination variation, shape change of target and partial occlusion.

Particle filter has already been extensively applied to video object tracking, however the traditional visual tracking approaches based on particle filter, which employ an experiential state transition equation and observation likelihood model predefined in advance, cannot satisfy the request of the real time complex situation video object tracking. In order to improve the robustness and stability as well as the computation efficiency of the video tracker based on particle filter, the adaptive state evolution equation and an online increment learning observation likelihood model are embedded into the particle filter, and the strategy for online self-adjusting the number of particle is adopted to enhance the computation efficiency. The experimental results show that the approach proposed in this paper not only track the moving object in the video accurately and effectively, but has nice robustness to the appearance variation caused by illumination and pose changes.

6-DOF (Degree of Freedom) head-tracking system based on CCD image processing is developed, and the measure errors of 6-DOFs are analyzed, which is caused by beacon accuracy and interdependence among 6-DOFs. In order to reduce the initial error caused by beacon accuracy, initial error correction program that the degrees return to zero at initial state is proposed. Then the initial errors are corrected without high-precision beacon. In addition, independent measurement program that the freedoms are extracted separately is proposed to avoid error transfer. Experimental results show that 6-DOF errors are reduced greatly after error correction and improved algorithm. In the azimuth angle range of -30°~30°, the standard deviation of azimuth angle α is 0.94°, the standard deviation of roll angle β and pitch angle γ in 0° position are 0.30° and 0.02° respectively, the standard deviation of position xo and yo in 0 cm position are 0.94 cm and 0.01 cm respectively, and the standard deviation of position zo in 112.31 cm position is 0.28 cm.

Adaptive optics systems with larger size require more subapertures in the Shack-Hartmann wavefront sensor, which makes it harder to compute the wavefront gradient in real time. In order to measure local gradients in large-size adaptive optics systems, an architecture based on cross correlation is proposed. Using the Gaussian spot as the reference template, the correlation between subaperture spots is calculated by means of two cascading filters, and then the local wavefront gradient of each subaperture is determined. Compared with the existing parallel processing approach in correlating-Hartmann sensor, the cost of our architecture will not increase greatly as subapertures multiplies, which makes it more suitable for adaptive optics systems of extreme size to retrieve gradients. Experimental results indicate that when implemented in Field Programmable Gate Array (FPGA) at 27 MHz, our design can obtain gradients of all subapertures in 8×8 Shack-Hartmann with latency no more than 1.2 μs, while the RMS error is less than 0.02 pixels, and the maximum error does not exceed 0.04 pixels. It can satisfy the requirements of low cost, real time and high accuracy for gradient detection in adaptive optics systems with large-size aperture.

When tomography is performed with strongly refracting field, refraction may cause sufficient bending of the probing rays. Ordinary reconstruction algorithm, based on the assumption of straight rays, does not yield accurate results. A new approach to curved-ray tomographic reconstruction of strongly refracting refractive-index distribution is proposed to realize a nonlinear inversion. A digital ray tracing method is designed for calculation of ray trajectory in inhomogeneous media. An initial estimate of the field is made by algebraic reconstruction technique. This estimate is then corrected by an iterative process in which computational ray tracing is used to generate deflection angles that are indirectly compared with the measured values. The reconstruction technique was tested by using simulated projection data of strongly refracting field for incomplete projection situation, and some reconstructed errors were discussed.

When detecting ballistic midcourse target with the Space-based Infrared System-low(SBIRS-Low), the background radiation of atmosphere or stellar is usually much stronger than that of target. Detection midcourse target is difficult. Based on design requirements of space-based infrared system-low, a method of detecting ballistic midcourse target with the space-based multi-band image system is inversed. In general, endmembers of target and atmosphere background are uncertain, and endmember of stellar radiation can been obtained by estimating the atmosphere absorbance. In this condition, a new algorithm based on OSP algorithm is proposed. Background endmembers matrix is made up of stellar endmember and atmosphere endmembers extracted with PCA method. Target endmembers is replaced with a unit vector. Computer simulations results show that the algorithm can effective suppress the background clutter of stellar and atmosphere. And midcourse target can been detected reliable.

The perfect theoretical calculation models of the satellite’s temperature field and infrared feature are established. First, according to the space locations of the satellite, the earth and the sun, the space heat flux that the satellite gets is calculated. Second, the surface temperature distribution of the satellite is got by solving the transient heat balance equations with the use of the Finite Element Method (FEM). And on the basis of the solved surface temperature distribution, taking the satellite as a point object, the spatial distribution of infrared radiation intensities in two IR bands is calculated. Finally, infrared radiation intensities in two IR bands of the satellite, the satellite’s body and solar panels are compared. The influence on the satellite’s infrared feature of the spatial heat flux is analyzed in detail. The research results have referential value on infrared detecting and recognition of spatial targets.

To intelligently evaluate the detectability of DSP early-warning satellite, a simulated scheme of system modeling for the process of satellite detecting is proposed based on the detection probability, the false alarm probability, the first instant when a target is captured and the minimum detection altitude. Through the assistance of STK, and the establishment of a series of models for the positioning parameters between satellite and missile, for the infrared radiation of missile target, the trajectory of missile, atmospheric background radiation and the noise of satellite detector, we constructed the simulated system for evaluating the detectability of DSP early-warning satellite. Some major findings were indicated when the system was applied in a simulation research. If the position relationship between satellite and missile remains constant and the weather condition is stable, the detectability of DSP early-warning satellite varies inversely as the detection probability of satellite and directly as the false alarm probability of satellite firstly, then keeps stable. If the detection probability of the satellite, the false alarm probability of the satellite and the position relationship between satellite and missile all remain constant, the detectability of the satellite within the wave band of 2.7 μm in clear weather is superior to that in cloudy weather. If the detection probability of the satellite and the false alarm probability of the satellite remain constant and the weather condition is stable, the detectability of the satellite for low latitude target is superior to its detectability for high latitude target.

In order to keep the regularity of edges or contour in the magnification of infrared image, an infrared image interpolation algorithm using Aliasing-free Contourlet Transform (AFCT) is proposed. Firstly, Aliasing-free Pyramidal Filter Banks (AFPFB) is combined with Directional Filer Banks (DFB) to construct AFCT. Then, a simple wavelet-based linear interpolation result is used as the initial estimate of the magnified image and fed in an iterative process. In the each iteration, the estimation of the interpolated image is viewed as a noisy version of the ideal high-resolution image and imposing AFCT to it. Subsequently, the denoising process is implemented via the sparsity constraint of the AFCT coefficients. Finally, the magnified infrared image is obtained after several iterations. Experiments show that for the test images, the iterative processing improves the Peak Signal to Noise Ratio (PSNR) with an average 0.837 dB. Moreover, the proposed algorithm obviously outperforms traditional methods based on bilinear interpolation or wavelet in terms of visual quality.

A tracking algorithm of infrared target is proposed by combining non-linear edge detection and Mean Shift method. The non-linear edge detection algorithm employs dual-window arithmetic operators that have the advantage of few calculation amount, high speed and good image quality and so on. The result of edge detection is binary images. Based on this information, the Mean Shift method is improved to implement target tracking. The tracking algorithm of improved Mean Shift combines the information of the local standard deviation calculation of the target area, describes the target based on the probability density function about gray value and the local standard deviation and selects cascade kernel function to calculate the target density that make up the shortage only using gray to describe the target features. Experimental results show that the edge of infrared target under complex background is detected clearly and infrared target is auto-tracked accurately.

It is important for the infrared thermal imaging system to increase the detection probability of small targets by resisting the interference of similar targets such as clouds, fog and objects on the ground. In order to decrease the false alarm rate of infrared surveillance system, a method for resisting of similar targets is put forward based on single infrared image. Firstly, the background noise is removed to enhance the SNR of infrared image. And then, according to head, tail and the length characters of targets, some potential targets are separated from the similar targets. The experiment result shows that this algorithm can decrease interference of similar targets and increase detection probability of infrared small target, and is very useful for small target detection.

A fusion algorithm of infrared and visible images was proposed based on Nonsubsampled Contourlet Transform (NSCT) and Pulse Coupled Neural Networks (PCNN). Firstly, two registered original images were decomposed by using NSCT separately, thus the low frequency subband coefficients and varieties of directional bandpass subband coefficients were obtained. Secondly, the selection principle of the low frequency subband coefficients was based on edges of images. The selection principle of the bandpass directional subband coefficients was improved by fusion method based on PCNN. Finally, the fused image was obtained by performing the inverse NSCT on the combined coefficients. The experimental results show that the proposed algorithm outperforms laplacian-based, wavelet-based and NSCT-based fusion algorithms.

A photonic crystal fiber for broadband dispersion-compensating with low nonlinearity in a wide wavelength range centered at 1 550 nm is designed, and the effects of the number of air-hole rings, the air-hole diameter and the lattice pitch on both dispersive and nonlinear properties are numerically described by using the vectorial beam propagation method. The results indicate that the inner six air-hole rings have stronger impact on the dispersive and the nonlinear properties. By optimizing and adjusting the diameters of the inner 4th to 6th air-holes and the lattice pitch, the proposed photonic crystal fiber can compensate (to within 0.5%) the dispersion of 190 times of its length of standard single mode transmission fiber over the entire 100nm band centered at 1 550 nm, while also can remain the nonlinear coefficient lower than 3 W-1?km-1 over the wide wavelength range.

To deal with the robust face recognition problem, a mixed model based on Hidden Markov Model (HMM) and Support Vector Machine (SVM) was proposed for HMM has good ability for time sequence modeling and SVM has excellent ability for classifying. Firstly, a sequence of overlapping sub-images was extracted from face image by using Discrete Cosine Transform (DCT) and Singular Value Decomposition (SVD). Then, the sequence which was extracted from training images was modeled by using HMM, and the output probability of each HMM for the training sequence had been considered as the input vector of SVM for its training. Finally, the output probability of each HMM for the testing sequences had been considered as the input vector of SVM for its testing. Experimental results on ORL and Yale face database demonstrate that the effectiveness and robustness of the proposed algorithm are better than traditional signal HMM and SVM algorithm.

Now the design and fabrication technology of large aperture SiC mirror assembly is the key point during the process of lightweight of the camera. The design process of a large aperture SiC mirror assembly is shown, which is the key structure of a lightweight camera, and a main flowchart of design techniques of a highly reliable, efficient and reasonable large aperture mirror assembly is given from the aspects of the principle prototype design, dynamic prototype design and ground test. Experimental results indicate that the 1st natural frequency of the SiC mirror assembly is 340Hz .There is 8.5% difference between the Finite Element Analysis (FEM) analysis results and the experimental results. At the same time，the difference of the mirror synthesis face figure error RMS was 0.009λ before and after the vibration experiment. All the experimental results show that the SiC mirror assembly satisfies the requirements of lightweight camera very well. Moreover, the design technology flowchart of the SiC mirror assembly is adaptive to the other large aperture mirror assembly design process. The result shows that it is a good method to improve the reliability of the large aperture mirror assembly and shorten the development periods.

According to the work principle of one aerial reconnaissance CCD camera, Descartes coordinate systems are properly build. Transformation of the coordinates is used to describe the relationship with two images which were taken in different times and scanning angles. Image affine transformation model with disturbance is derived by co-line principle. Then, a simulation analysis is given. Results of simulation indicate that volatility of image overlap rate in the direction of heading and side is determined by pitching and roll angle disturbance separately, and it can be reduced into a certain range by compensating the two attitude disturbance. Finally, it’ll meet the demand of image mosaicing. Flying test result has proved this conclusion effectively. It can supply a theory basis for the overlap rate and compensation mechanism design of the same species camera, and it also can simplify the registration process of the image mosaicing.

Factors which affect the imaging quality of CMOS image sensor LUPA-4000 were investigated by analyzing the power supplies and the driving timing. Considering these factors, a CMOS imaging system with adjustable power supplies and driving timing was designed and implemented based on FPGA and ARM. Seven power supplies needed to be adjusted during LUPA-4000’s working were designed by means of the combination of adjustable resistances and corresponding power chips. Parameters used in the adjustment of LUPA-4000’s driving timing were set in PC firstly, and then transported to FPGA through Ethernet based on ARM. Actual tests showed that the imaging system can be operated and expanded flexibly, and can adjust all kinds of parameters in LUPA-4000’s driving timing precisely. Three main topics are discussed in this imaging system, including design of adjustable power supplies, methods of parameters’ setting and transmitting, and FPGA implement of LUPA-4000’s driving timing.

Multiple optical payloads are integrated on an opto-electronic platform of tethered balloon for the surveillance of Expo Site of Shanghai. These optical payloads are exigent for the working temperature in order to get high quality image. Thermal models of payloads and the platform itself are built by Sinda/Fluint software. Based on the thermal model, a specific thermal design is worked out for the payload and the platform. The approach consists of both active and passive thermal control measures. It is showed that the thermal design is satisfactory by software simulation.

To take control of the propagation direction over the electromagnetic wave, or even over the propagation direction of wave, including light-wave, is a promising research target. The controllable negative refractive index materials were researched, and by using the changes of refractive index, the goal of controlling wave direction could be achieved. Based on the analysis about the feature of generating negative magnetic permeability by Double Split-ring Resonator (DSRR) array, the magnetic permeability and the final changes of refractive index caused by the states swaying from Double Part-closed-ring Resonator (DPCRR) array were researched, and different characteristics produced by opening and closing were found. Then we designed an array controllable program in the DPCRR array. Some experiments were carried out about the two designed structure. Compared with the theoretical structures, the controllable negative refractive index structure designed in this thesis was validated. The final result proves that this structure can make the two propagation directions of some microwave-bands become controllable. At the same time, the experimental result demonstrates feasibility of using the changes of refractive index to control the direction of wave transmission.

The basic reason of low retrieval rate of contourlet transform texture image retrieval system lies in the coefficients shift sensitive characters and poor local time-frequency character. In order to overcome the defects of contourlet transform, a Contourlet-2.3 transform based on mapping is proposed, and the shift invariant character is proved. A texture image retrieval system based on the new transform is proposed in which the feature vectors are formed by cascading the energy and standard deviation of each sub-band in the transform domain, and Canberra distance is used as similarity metric. Experimental results show that the image retrieval system is superior to that of the original contourlet transform, Contourlet-2.3 and contourlet transform based on mapping with the same length of feature vectors, retrieval time and memory.

The existing fisheye correction methods have only utilized single image and are lack of enough information to correct distortion errors due to some factors, such as camera shaking and environment disturbing. An automatic and accurate fisheye image correction algorithm is presented based on the matched feature points between two fisheye images, which can automatically correct the shooting errors and preserve the vertical invariance of corrected images by using the estimated pitch angles. Experiment shows that the proposed method can eliminate the ghost of image stitched caused by the distortion errors, and enhance the effect of the stitched images.